Endoscopic ultrasound: accuracy in staging superficial carcinomas of the esophagus.

Background: Endoscopic procedures may represent an alternative to esophagectomy for superficial neoplasms of the esophagus (T1m/T1sm), but they are considered curative only in case of no lymph node involvement. Endoscopic ultrasound (EUS) is the most accurate method to define both T and N staging of esophageal carcinoma. Aims of the study were to assess the staging accuracy of EUS in superficial lesions (T1m and T1sm) of patients who were candidates for esophagectomy or local endoscopic resection and to establish which variables (site of neoplasm, histologic type, macroscopic appearance) can affect the accuracy of EUS in distinguishing between T1m and T1sm lesions.

Methods: The study population consisted of 55 patients with superficial carcinoma of the esophagus who underwent EUS (October 2002 to January 2007). Endoscopic ultrasound features were compared with findings from surgical specimens or samples obtained at mucosectomy.

Results: There were 33 patients with adenocarcinoma (60%), which developed on Barrett's esophagus in 27 cases, 21 patients (38%) with squamous cell carcinoma, and 1 (2%) with lymphoepithelial-like carcinoma. All lesions were confirmed as T1 on pathology. Of the 22 (40%) T1m lesions on EUS, 19 (86%) were confirmed as T1m on pathology; of the 33 T1sm on EUS, 22 (66%) were confirmed as T1sm. Positive predictive value of EUS for invasion of the submucosa was 67%, negative predictive value 86%, sensitivity 88%, specificity 63%, and diagnostic accuracy 75%. The accuracy of EUS in evaluating lymph node metastases was 71%, with a negative predictive value of 84%. Endoscopic ultrasound accuracy in differentiating mucosal from submucosal lesions increased from the lower esophagus or gastroesophageal junction to the mid and upper esophagus (71%, 76%, and 100%, respectively; not significant). As for the histologic type, accuracy was 70% for adenocarcinoma and 81% for squamous cell carcinoma, (not significant); for lesions detected as type 0-IIa (13 patients), accuracy was 100%; for type 0-I lesions (23 patients), accuracy was 70% (p = 0.03).

Conclusions: Despite difficulties in differentiating mucosal from submucosal lesions, even with 20-MHz miniprobes, EUS remains an extremely valuable tool when nonsurgical treatments are considered. Its staging accuracy depends on site and macroscopic appearance of the neoplasm.